CityUHK Data Science PhD Student Shines at NeurIPS 2024 AIGB Bidding Competition
18 DEC 2024
The Department of Data Science (DS) at City University of Hong Kong (CityUHK) is proud to announce the exceptional achievement of its student in the prestigious competition NeurIPS 2024 AIGB Track: Learning Auto-Bidding Agents with Generative Models, which took place on 14 December 2024. A collaborative team, KGAB, consisting of members from CityUHK’s AML Lab and Kuaishou, secured 1st place overall out of 1,500+ teams.
The team included GAO Jingtong, a PhD candidate from DS AML Lab, as the key contributor, under the guidance of Professor ZHAO Xiangyu and in collaboration with Kuaishou researchers. Their innovative solution successfully addressed the competition’s challenges, earning them the top spot.
This year’s competition brought together over 1,500 teams from leading universities, research labs, and industry leaders, requiring participants to develop generative bidding agents capable of navigating strict budget and CPA constraints. KGAB’s winning approach featured three key innovations:
- Constraint-Aware RTG Optimization
A specialized Return-to-Go (RTG) framework that incorporated real-world constraints, ensuring effective and practical bidding strategies. - RTG-Driven Exploration
A novel mechanism to improve exploration and identify strategies that maximize long-term value while maintaining safety and stability. - Offline Simulation Environment
A customized simulation dataset to rigorously test and validate strategies in a controlled setting before deployment.
KGAB’s creative use of generative models and exploration techniques demonstrated a strong understanding of both theoretical and practical aspects of decision intelligence.
This achievement underscores DS’s leadership in AI and data science research and its commitment to advancing innovation through collaboration between academia and industry.